Few figs for frugivores: Riparian fig trees in Zimbabwe may not be a dry season keystone resource

Most plants flower and fruit at times of year when probabilities of pollination and seedling establishment are high. Fig trees ( Ficus spp.) are often considered as key stone resources for vertebrate frugivores, in part because of year-round fig pro -duction. This unusual fruiting phenology results in the maintenance of fig wasp populations, but in seasonal environments this means fruiting occurs during periods when the chances of seedling establishment are low. Under these circumstances, se lection is expected to favour any individuals that reduce or eliminate fruiting at these times. Here, we describe a large-scale survey of the extent of dry season fruiting by three riparian Ficus species in Gonarezhou National Park, Zimbabwe. Few trees of two monoecious species, F. sycomorus and F. abutilifolia, had figs, and most crops of F. sycomorus were far smaller than the trees were capable of producing. Large stands of the dioecious F. capreifolia were present, but fig densities were low and no mature female (seed containing) figs were recorded. Even though fig trees may have been the only species

Their keystone status reflects the importance of figs in the diets of many tropical mammals and birds, with more vertebrates recorded as eating figs than any other fleshy fruits (Shanahan, So, Compton, & Corlett, 2001) and an increasing realisation that the insects associated with the figs are themselves important in the diet of birds (Mackay, Gross, & Rossetto, 2018 year when few other fruits are available (Foster, 2014;Lambert & Marshall, 1991;O'Brien et al., 1998;. Many tropical and subtropical trees display a sub-annual, synchronised flowering pattern (Kinnaird, 1992, Chapman, Wrangham, & Chapman, 1994, but see Kattan & Valenzuela, 2013). This contrasts with the year-round fruiting displayed by many fig trees, which is seen as being a particularly significant trait because it means that they can support frugivore populations through periods of shortage when little other food is available (van Schaik, Terborgh, & Wright, 1993;Tweheyo & Lye, 2003;Walther, Geier, Lien-Siang Chou, & Bain, 2018). and their year-round fruiting is linked to their protogynous inflorescences and dependence on these short-lived insects for pollination (Janzen, 1979). The year-round fruiting pattern of many monoecious Ficus species is often associated with between-tree fruiting asynchrony, but within-tree synchrony (Bronstein, Gouyon, Gliddon, Kjellberg, & Michaloud, 1990). Within-tree fruiting synchrony forces outcrossing and may aid pollinator attraction (Janzen, 1979) (Kjellberg & Maurice, 1989;Patel, 1996). Their more  (Garcia, Hong, & Ellis, 2005; Vázquez-Yanes, Rojas-Aréchiga, Sánchez-Coronado, & Orozco-Segovia, 1996) and the trees are likely to be investing resources into reproduction at times of the year when opportunities for seeds to survive and seedlings to successfully establish are minimal. Species inhabiting arid environments that produce seeds during the dry season are therefore less likely to generate new plants than at other times of year. Fig production during the dry season can also contribute to water stress on the trees (Patiño, Herre, & Tyree, 1994). These factors suggest that natural selection should favour individuals that reduce their reproductive investment at such times (Kjellberg & Maurice, 1989) Views on what constitutes a keystone species vary, but definitions often include having a high ecological significance at large spatial scales and interacting with numerous other species (Jordán, 2009;Mills, Soulé, & Doak, 1993).

| Study sites
Gonarezhou National Park is situated in the Lowveld of south-east Zimbabwe, approximately 21° south of the Equator. The climate of the region is strongly seasonal, with a long dry season punctuated by summer rains (Torrance, 1981). The Runde (Lundi) River is the only major river in the northern section of the park, where it runs for about 74 km from the western park boundary through to the junction with the Save (Sabi) River, close to the border with Mozambique.
The Runde is ephemeral, with above-ground water restricted to iso- Camp where low cliffs border the river. These changes in substrate divide the river into five discrete areas, which (from west to east) we refer to as the Bridge, Chipinda, Central, Chitove and Junction sections. Consequently, our riverside censuses, which took place in July and

| Study plants
August 1994, will have covered almost all the fig trees in the area.
Ficus sycomorus is a medium-sized monoecious tree that has a preference for riverine environments in drier areas (Burrows & Burrows, 2003). Its large figs (40 mm diameter) are produced on leafless branches from the trunk and major branches. The figs are pollinated by Ceratosolen arabicus Mayr and fed on by a wide range of birds, fruit bats and primates when ripe (Brain, 1988;.
Ficus abutilifolia is a monoecious shrub or small tree. It is a 'rock-splitter', restricted to rocky areas. Its pollinator is Elisabethiella comptoni Wiebes (Berg & Wiebes, 1992), and its figs are located in the leaf axils.
They reach 20 mm in diameter and are eaten by birds and mammals

| Sampling
The Runde River and adjacent sections of major tributaries were sur-

| RE SULTS
Our census recorded a total of 255 large F. sycomorus (Table 1). The trees were present at relatively low densities along most of the length of the river and for short distances up the major tributaries, but were absent from the two areas with rocky substrates (Figure 2).
They averaged about five trees per kilometre along the nonrocky stretches of river, with the highest concentrations around junctions with the tributaries and at the sides of some of the larger persistent pools. Mature individuals were readily identified by the presence of active or old fruit-bearing branches. There was little evidence of recent recruitment, although five small trees were present and groups of around 20 small saplings were present beneath two of the mature trees. These saplings were unlikely to survive for long as they were situated in the bed of the river, in contrast to all the mature trees, which were on the riverbanks or on stabilised sand bars.
In contrast to F. sycomorus, F. abutilifolia was only found in the rocky Chipinda and Chitove stretches of the river. The Chipinda (more westerly) rocky section was by far the longer of the two areas and supported most of the 808 F. abutilifolia trees considered sufficiently large to have potentially borne figs (Table 1)  Taken as a group, fig trees were present along the whole length of the river, but the numbers of potentially fruiting trees were concentrated along certain sections due to the clumped distributions of two of the three species. These localised relatively high densities of fig trees did not lead to large numbers of mature figs being available, however. Fruiting frequencies were higher in F. sycomorus, with figs present on about 40% of the trees, but only a proportion of these had mature figs available to be eaten. Furthermore, crop sizes were generally extremely small for such large trees, with most crops numbering <100 figs (Table 1 and Figure 3a) Based on the censuses of F. sycomorus and F. abutilifolia and sampling of the F. capreifolia populations, we estimate that for all three species combined there were less than 5,000 figs that were mature and waiting to be eaten on the 6,000 or so trees present along the 74 km of river (Table 2). Development of most of the crops of F. sycomorus and F. abutilifolia was sufficiently synchronised to prevent self-pollination, although a small number of trees did have unpollinated figs while simultaneously releasing pollinator fig wasps (Table 3).   (Garcia et al., 2005;Vázquez-Yanes et al., 1996). There are also few receptive figs on the trees at this time, so reproductive success via dispersal of their pollen will also be low. The low dry season fruiting frequencies we observed may therefore be adaptive, but some individuals nonetheless continued to fruit, rather than preserving their resources entirely by ceasing to produce figs. It may be that the costs involved in fruiting during one part of the year have minimal influence on the resources available to support fruiting at other times. Alternatively, there may be an element of local maladaptation in the trees' dry season phenologies that reflects the long-distance gene flow exhibited by species such as F. sycomorus. Phenotypic plasticity is nonetheless present in the congeneric F. thonningii, which varies its fruiting in response to water availability in Zimbabwe (Damstra, Richardson, & Reeler, 1996). There will also be rewards for those individuals producing fig wasps towards the end of the dry season, because they have the potential to pollinate figs which subsequently produce seeds when conditions are more favourable.

| D ISCUSS I ON
Alternatively, their low dry season fruiting investment may be a direct response to environmental conditions, particularly the availability of water, at least in F. sycomorus. Water stress can limit fruit production by fig trees (Patiño et al., 1994), and it was noticeable that the few individuals of this species that had larger crops were located next to the remaining major pools in the riverbed.
Elsewhere in Africa, riparian fig trees such as F. sycomorus can be much more abundant than in our study area (Makishima, 2005) and they even form virtual monocultures along rivers in Kruger Park, South Africa (Adams & Snode, 2013;Bonaccorso et al., 2014). The abundance of riparian fig trees therefore varies greatly, but they can retain ecological significance in deserts even when at low densities (Ahmed et al., 2009;Brain, 1988;Wharton, Tilson, & Tilson, 1980).

| CON CLUS ION
Our results emphasise that a blanket assumption of keystone status

ACK N OWLED G EM ENTS
The extensive field work would not have been possible without the leadership of M. J. Burke (Metropolitan Police Service, London), the dedicated work of Raleigh International support staff and venturers, and the staff of Gonarezhou National Park. We thank the three referees for useful comments that improved the manuscript.

CO N FLI C T O F I NTE R E S T
The authors declare no competing interests.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data supporting the findings of this study are available from the corresponding author upon reasonable request.